Modelling of transmembrane pressure using slot/pore blocking model, response surface and artificial intelligence...

Area of Science:

• Membrane Science and Technology
• Chemical Engineering
• Artificial Intelligence in Process Engineering

Background:

• Accurate prediction of transmembrane pressure (TMP) is crucial for optimizing membrane filtration processes, especially for deformable oil droplets.
• Existing empirical and statistical models may not fully capture the complex interactions influencing TMP in such systems.
• Deformable oil droplet treatment presents unique challenges in membrane fouling and pressure dynamics.

Purpose of the Study:

• To investigate and compare the predictive capabilities of empirical, statistical (Response Surface Methodology - RSM), and machine learning (Artificial Neural Network - ANN) methods for transmembrane pressure (TMP).
• To assess the prediction accuracy and generalization of these models for crude oil and Tween-20 oil-in-water emulsions using an oscillating slotted pore membrane.
• To identify key operational parameters influencing TMP and determine optimal conditions for minimizing TMP.

Main Methods:

• Utilized 87 experimental runs with permeate flux, shear rate, and filtration time as input features.
• Developed and compared Artificial Neural Network (ANN) models against Response Surface Methodology (RSM) and the empirical Slot-Pore Blocking Model (SBM).
• Employed line plots, contour plots, and sensitivity analysis to interpret model predictions and parameter influence.

Main Results:

• The ANN model, with 10 hidden neurons, demonstrated superior accuracy in approximating TMP for both crude oil and Tween-20 compared to RSM and SBM.
• Analysis revealed linear relationships between TMP and flux rate/filtration time, and an inverse relationship between TMP and shear rate.
• Sensitivity analysis identified flux rate as the most influential parameter on TMP, followed by filtration time and shear rate.

Conclusions:

• Artificial Neural Network (ANN) modeling is a highly effective tool for accurate TMP prediction in oscillating slotted pore membrane systems treating deformable oil droplets.
• The findings provide insights into parameter optimization, suggesting higher shear rates and lower flux rates/filtration times minimize TMP.
• This study supports the use of ANN for enhanced process design and scale-up in membrane-based oil-in-water separation.